Loris Marchal

Scheduling distributed applications: the SimGrid simulation framework

Since the advent of distributed computer systems an active field of research has been the investigation of scheduling strategies for parallel applications. The common approach is to employ scheduling heuristics that approximate an optimal schedule. Unfortunately, it is often impossible to obtain analytical results to compare the efficacy of these heuristics. One possibility is to conducts large numbers of back-to-back experiments on real platforms. While this is possible on tightly-coupled platforms, it is infeasible on modern distributed platforms (i.e. Grids) as it is labor-intensive and does not enable repeatable results. The solution is to resort to simulations. Simulations not only enables repeatable results but also make it possible to explore wide ranges of platform and application scenarios. In this paper we present the SimGrid framework which enables the simulation of distributed applications in distributed computing environments for the specific purpose of developing and evaluating scheduling algorithms. This paper focuses on SimGrid v2, which greatly improves on the first version of the software with more realistic network models and topologies. SimGrid v2 also enables the simulation of distributed scheduling agents, which has become critical for current scheduling research in large-scale platforms. After describing and validating these features, we present a case study by which we demonstrate the usefulness of SimGrid for conducting scheduling research.

Scheduling Concurrent Bag-of-Tasks Applications on Heterogeneous Platforms

Scheduling problems are already difficult on traditional parallel machines, and they become extremely challenging on heterogeneous clusters. In this paper, we deal with the problem of scheduling multiple applications, made of collections of independent and identical tasks, on a heterogeneous master-worker platform. The applications are submitted online, which means that there is no a priori (static) knowledge of the workload distribution at the beginning of the execution. The objective is to minimize the maximum stretch, i.e., the maximum ratio between the actual time an application has spent in the system and the time this application would have spent if executed alone. On the theoretical side, we design an optimal algorithm for the offline version of the problem (when all release dates and application characteristics are known beforehand). We also introduce a heuristic for the general case of online applications. On the practical side, we have conducted extensive simulations and MPI experiments, showing that we are able to deal with very large problem instances in a few seconds. Also, the solution that we compute totally outperforms classical heuristics from the literature, thereby fully assessing the usefulness of our approach.

VoroNet: A scalable object network based on Voronoi tessellations

In this paper, we propose the design of VoroNet, an object-based peer to peer overlay network relying on Voronoi tessellations, along with its theoretical analysis and experimental evaluation. VoroNet differs from previous overlay networks in that peers are application objects themselves and get identifiers reflecting the semantics of the application instead of relying on hashing functions. This enables a scalable support for efficient search in large collections of data. In VoroNet, objects are organized in an attribute space according to a Voronoi diagram. VoroNet is inspired from the Kleinbergs small-world model where each peer gets connected to close neighbours and maintains an additional pointer to a long-range neighbour. VoroNet improves upon the original proposal as it deals with general object topologies and therefore copes with skewed data distributions. We show that VoroNet can be built and maintained in a fully decentralized way. The theoretical analysis of the system proves that routing in VoroNet can be achieved in a poly-logarithmic number of hops in the size of the system. The analysis is fully confirmed by our experimental evaluation by simulation.

Pipelining broadcasts on heterogeneous platforms

In this paper, we consider the communications involved by the execution of a complex application, deployed on a heterogeneous platform. Such applications extensively use macrocommunication schemes, for example, to broadcast data items. Rather than aiming at minimizing the execution time of a single broadcast, we focus on the steady-state operation. We assume that there is a large number of messages to be broadcast in pipeline fashion, and we aim at maximizing the throughput, i.e., the (rational) number of messages which can be broadcast every time-step. We target heterogeneous platforms, modeled by a graph where resources have different communication and computation speeds. Achieving the best throughput may well require that the target platform is used in totality: we show that neither spanning trees nor DAGs are as powerful as general graphs. We show how to compute the best throughput using linear programming, and how to exhibit a periodic schedule, first when restricting to a DAG, and then when using a general graph. The polynomial compactness of the description comes from the decomposition of the schedule into several broadcast trees that are used concurrently to reach the best throughput. It is important to point out that a concrete scheduling algorithm based upon the steady-state operation is asymptotically optimal, in the class of all possible schedules (not only periodic solutions).

Centralized versus Distributed Schedulers for Bag-of-Tasks Applications

Multiple applications that execute concurrently on heterogeneous platforms compete for CPU and network resources. In this paper, we consider the problem of scheduling applications to ensure fair and efficient execution on a distributed network of processors. We limit our study to the case where communication is restricted to a tree embedded in the network, and the applications consist of a large number of independent tasks (Bags of Tasks) that originate at the trees root. The tasks of a given application all have the same computation and communication requirements, but these requirements can vary for different applications. The goal of scheduling is to maximize the throughput of each application while ensuring a fair sharing of resources between applications. We can find the optimal asymptotic rates by solving a linear programming problem that expresses all necessary problem constraints, and we show how to construct a periodic schedule from any linear program solution. For single-level trees, the solution is characterized by processing tasks with larger communication-to-computation ratios at children with larger bandwidths. For multilevel trees, this approach requires global knowledge of all application and platform parameters. For large-scale platforms, such global coordination by a centralized scheduler may be unrealistic. Thus, we also investigate decentralized schedulers that use only local information at each participating resource. We assess their performance via simulation and compare to an optimal centralized solution obtained via linear programming. The best of our decentralized heuristics achieves the same performance on about 2/3 of our test cases but is far worse in a few cases. Although our results are based on simple assumptions and do not explore all parameters (such as the maximum number of tasks that can be held on a node), they provide insight into the important question of fairly and optimally scheduling heterogeneous applications on heterogeneous grids.

STEADY-STATE SCHEDULING ON HETEROGENEOUS CLUSTERS

This paper considers steady-state scheduling techniques for heterogeneous systems, such as clusters and grids. The use of steady-state scheduling is advocated to solve a variety of important problems, which would be too difficult to tackle with the objective of makespan minimization. Several examples are given, namely master-slave tasking, mixed task and data parallelism, and pipelined macro-communications (scatter, broadcast, multicast). For each example, both the advantages and the limitations of the approach are discussed.

Centralized versus distributed schedulers for multiple bag-of-task applications

Multiple applications that execute concurrently on heterogeneous platforms compete for CPU and network resources. In this paper, we consider the problem of scheduling applications to ensure fair and efficient execution on a distributed network of processors. We limit our study to the case where communication is restricted to a tree embedded in the network, and the applications consist of a large number of independent tasks that originate at the trees root. The tasks of a given application all have the same computation and communication requirements, but these requirements can vary for different applications. Each application is given a weight that quantifies its relative value. The goal of scheduling is to maximize throughput while executing tasks from each application in the same ratio as their weights. We can find the optimal asymptotic rates by solving a linear program that expresses all necessary problem constraints, and we show how to construct a periodic schedule. For single-level trees, the solution is characterized by processing tasks with larger communication-to-computation ratios at children with larger bandwidths. For multi-level trees, this approach requires global knowledge of all application and platform parameters. For large-scale platforms, such global coordination by a centralized scheduler may be unrealistic. Thus, we also investigate decentralized schedulers that use only local information at each participating resource. We assess their performance via simulation, and compare to a centralized solution obtained via linear programming. The best of our decentralized heuristics achieves the same performance on about two-thirds of our test cases, but is far worse in a few cases. While our results are based on simplistic assumptions and do not explore all parameters (such as buffer size), they provide insight into the important question of fairly and optimally co-scheduling heterogeneous applications on heterogeneous grids

Broadcast trees for heterogeneous platforms

In this paper, we deal with broadcasting on heterogeneous platforms. Typically, the message to be broadcast is split into several slices, which are sent by the source processor in a pipeline fashion. A spanning tree is used to implement this operation, and the objective is to find the tree, which maximizes the throughput, i.e. the average number of slices sent by the source processor every time-unit. We introduce several heuristics to solve this problem. The good news is that the best heuristics perform quite efficiently, reaching more than 70% of the absolute optimal throughput, thereby providing a simple yet efficient approach to achieve very good performance for broadcasting on heterogeneous platforms.

Steady-State Scheduling of Multiple Divisible Load Applications on Wide-Area Distributed Computing Platforms

Divisible load applications consist of an amount of data and associated computation that can be divided arbitrarily into any number of independent pieces. This model is a good approximation of many real-world scientific applications, lends itself to a natural master-worker implementation, and has thus received a lot of attention. The critical issue of divisible load scheduling has been studied extensively in previous work. However, only a few authors have explored the simultaneous scheduling of multiple such applications on a distributed computing platform. We focus on this increasingly relevant scenario and make the following contributions. We use a novel and more realistic platform model that captures some of the fundamental network properties of grid platforms. We formulate the steady-state multi-application scheduling problem as a linear program that expresses a notion of fairness between applications. This scheduling problem is NP-complete and we propose several heuristics that we evaluate and compare via extensive simulation experiments. Our main finding is that some of our heuristics can achieve performance close to the optimal and we quantify the trade-offs between achieved performance and heuristic complexity.

Scheduling divisible loads with return messages on heterogeneous master-worker platforms

In this paper, we consider the problem of scheduling divisible loads onto an heterogeneous star platform, with both heterogeneous computing and communication resources. We consider the case where the workers, after processing the tasks, send back some results to the master processor. This corresponds to a more general framework than the one used in many divisible load papers, where only forward communications are taken into account. To the best of our knowledge, this paper constitutes the first attempt to derive optimality results under this general framework (forward and backward communications, heterogeneous processing and communication resources). We prove that it is possible to derive the optimal solution both for LIFO and FIFO distribution schemes. Nevertheless, the complexity of the general problem remains open: we also show in the paper that the optimal distribution scheme may be neither LIFO nor FIFO.